Infrastructure and cloud 8 min

Leaving SaaS: the real three-year cost for an SME

Leaving SaaS only pays off if the full cost of self-hosting over three years (server, migration, operation, backups, human hours) stays clearly below that of the subscriptions over the same period. The calculation is done tool by tool: software billed per user is first in line, while critical tools with no team to maintain them stay on SaaS.

Contents9
  1. 01How to tell whether leaving is worth it
  2. 02Step 1: the inventory, tool by tool
  3. 03Step 2: the SaaS cost over three years
  4. 04Step 3: the self-hosted cost over three years
  5. 05Step 4: compare tool by tool, not as a whole
  6. 06What the calculation does not see
  7. 07Honest limits
  8. 08Where to start
  9. 09Frequently asked questions

In short

  • The calculation compares two full costs over three years, the period over which a migration has time to pay for itself.

  • The inventory of subscriptions can already save money, before any migration.

  • On the SaaS side, two traps: the bill for a tool paid per user follows your hiring, and the monthly price is not the annual commitment price.

  • The calculation does not see everything: where the data is stored, reversibility or the human risk also count.

How to tell whether leaving is worth it

Leaving SaaS is not a matter of principle, it is a subtraction. On one side, what you will pay in subscriptions over three years if nothing changes. On the other, what self-hosting will cost over the same period, including the time of the people who will keep it running. If the gap is small, you stay put: migration has a cost and a risk, and a well-chosen subscription remains a good deal.

Three years, because that is long enough for a migration to pay for itself, and because beyond that, vendors’ prices and the company’s needs both change too much for the calculation to hold.

Our own server runs a dozen services, including the Twenty CRM, which we have operated since early 2025, n8n automation, Matomo analytics, electronic signature and invoicing. What follows is the method we apply before recommending anything to a client, in four steps:

The method in four steps

  1. The inventory

    Each subscription is listed, with its real use and its billing method.

  2. The SaaS cost

    Subscriptions are projected over three years, ancillary fees included.

  3. The self-hosted cost

    Server, licences, migration, operation and backups are counted over three years, human hours included.

  4. The comparison

    Tools are compared one by one, then sorted into three groups.

It is also where our Leaving SaaS offer begins: an audit of your subscriptions and their real cost over three years, before any migration.

Step 1: the inventory, tool by tool

In an SME of 50 people, the exact list of subscriptions is not always kept up to date: they may be paid on several cards, by several departments, some in dollars. Start with a table with, for each tool:

  • the name and the real use (who uses it, how many times a week);
  • the billing model: per user, by volume (contacts, sends, executions) or flat fee;
  • the number of licences paid for and the number actually used;
  • the renewal date and the commitment period;
  • extra charges: mandatory set-up, paid modules, overage.

This table can save money before any migration: licences for people who have left, duplicates between two teams, a module that was never switched on.

Step 2: the SaaS cost over three years

For each tool, project three years, not one. The calculation:

three-year SaaS cost = (price per licence × number of licences × 36 months) + set-up fees + modules + foreseeable overage charges

Two traps come up often. The first is growth: a tool billed per user follows your hiring. If you go from 50 to 70 people, the bill follows. The second is the difference between the price shown with an annual commitment and the monthly price.

An example of a public price, to show what the line looks like: HubSpot Sales Hub’s pricing page showed the Professional plan in US dollars, with several payment options.

$90

per licence per month: the starting price of the Professional plan

$1,500

mandatory set-up fee, charged once, on top of the licence price

Source: HubSpot Sales Hub pricing page, read in October 2026

These prices change: read them again on the day you do your calculation, and note the date next to the figure.

Do not mix currencies. Convert at the day’s rate and note it in the table.

Step 3: the self-hosted cost over three years

Many comparisons stop at the price of the server. The real cost has five lines.

The server and what surrounds it. A virtual private server or a dedicated server, a domain name, backup storage outside the server, possibly an email-sending service. These lines, which form the basis of the infrastructure foundation, are fixed whatever the number of users: that is the whole point.

Software licences. Many open-source tools are free when self-hosted and paid in the version hosted by the vendor. Twenty’s pricing page, for example, offered in October 2026 a cloud billed per user, and stated that the open-source core can be self-hosted without a licence, with single sign-on (SSO) and row-level permissions requiring the paid Organization plan. Check the licence of each tool, and the features reserved for the paid version.

The migration. Data export, cleaning, import, reconfiguration of integrations, training. It is a one-off expense, but it has to be recouped within the three years.

Operation. Updates, monitoring, certificate renewal, incident response. Count it in hours per month, multiplied by a real hourly cost, internal or invoiced. If nobody in the company knows how to do it, it is a managed-services contract, and it goes into the calculation.

Backups and their tests. A backup that has never been restored is worth nothing. The time spent on a restore exercise is part of the cost; we come back to it in our article on backing up a server of your own and what really breaks.

The calculation:

three-year self-hosted cost = (server + storage + ancillary services) × 36 + migration + (operating hours per month × hourly cost × 36) + any licences

Step 4: compare tool by tool, not as a whole

The overall result “we save X” hides what matters. Compare line by line, and sort the tools into three groups.

The good candidates. Tools billed per user, used by many people, whose self-hosted equivalent is mature. This is where the gap is widest and where it grows with the team. Our article on the tools to self-host first details the order we recommend.

The cases to discuss. Cheap tools, little-used tools, or tools whose open equivalent needs a lot of configuration. There is sometimes a saving, but it does not pay for the risk.

What you do not leave. The tools on which the risk of an outage weighs more than the saving.

At the two extremes, this gives:

The good candidates

  • The CRM

  • Automation

  • Web analytics

  • Electronic signature

  • Invoicing

What you do not leave

  • The company’s email and office suite

  • Statutory accounting, if your accountant works in a specific tool

  • Payroll

  • Any critical tool for which you have neither a team nor a provider able to step in on a Sunday

What the calculation does not see

Some criteria cannot be put into euros, and yet they weigh in the decision.

  • Where the data is located. Hosting it yourself or with a European host changes the answer to the question “where is our customer data?”. It is a real argument for some sectors, which we develop in our article on dependence on the American cloud.
  • Reversibility. An open-source tool whose database you hold is easier to leave than a SaaS whose export is partial.
  • Tools that talk to each other. On the same server, the CRM, automation and invoicing connect without going through three paid APIs.
  • Human risk. If only one person knows how to run the server, their departure is a risk. Operating documentation is not a luxury.

Honest limits

Self-hosting is not the right choice when the company has neither in-house skills nor a managed-services budget, when the SaaS tool is cheap and well used, or when compliance requires certifications that you cannot obtain on your own.

Watch out

Operating hours are never zero

Nor is self-hosting the right choice if the three-year gap exists only because you counted zero hours of operation. If your calculation arrives at that result, do it again.

Where to start

Do the inventory from step 1, even if you do not intend to migrate anything: it is the step that costs the least. Then put figures on two or three tools billed per user, not the whole stack. If you want this calculation done on your own subscriptions, our Leaving SaaS offer starts with this audit, and if the gap does not justify the migration, we tell you so. The other infrastructure choices, from which tools to leave first to backups, are gathered in our Infrastructure and cloud section.

Frequently asked questions

Is self-hosting profitable for a company?

Often for tools billed per user and used by many people, rarely for email or payroll. It all depends on what your subscriptions cost, on the growth of your team and on the hours of operation, which are never zero. The only reliable answer is the tool-by-tool calculation over three years: if the gap is thin, the subscription remains the better deal.

SaaS, on-premise or cloud: which hosting model should you choose?

SaaS is software rented and run by its vendor. On-premise means software installed on your own servers, in your premises. In between, you can rent a server from a host and install your tools on it: that is what we do. The choice is made tool by tool, not for the whole company, and nothing stops you combining all three.

Should you migrate everything at once?

No. You start with one or two tools with a big gap and low risk, run them for a few months, then decide what comes next. Each tool switches over one after another, and the old subscription stays active until the new tool has proved itself. A migration in one block multiplies the points of failure and makes going back difficult.

What happens if the server goes down?

That is the right question, and it is prepared before the migration: backups outside the server, a restore already tested, and a person or a provider who knows how to step in, under an agreed on-call arrangement. Without these three elements, do not migrate. With them, an outage becomes a recoverable incident, and its cost also goes into the three-year calculation.

Sources

  1. HubSpot Sales Hub’s pricing page (new tab)hubspot.com
  2. Twenty’s pricing page (new tab)twenty.com
Source 1

HubSpot Sales Hub’s pricing page

Read on hubspot.com(new tab)
Source 2

Twenty’s pricing page

Read on twenty.com(new tab)

Jim · founder of Brain Plus AI

Also in infrastructure and cloud

  • Practical guide7 min

    Backing up your own server: what really breaks

    Everyone knows the 3-2-1 rule, but what loses data is more mundane: a backup never restored, a forgotten volume, a global clean-up command run in production. Here is what we learned from operating our own server.

  • Practical guide7 min

    Self-hosting: which tools to move out of SaaS first

    Not all subscriptions are equal when it comes to self-hosting: some open-source tools are mature and easy to run, others need a real team. Here is the order we recommend, and what stays on SaaS.

  • Stepping back11 min

    Sovereign cloud: can you do without the American giants?

    European providers now hold only 15% of their own cloud market. Full sovereignty is out of reach; selective sovereignty starts with sorting your data and planning an exit.

WhatsApp